Benchmark
- Messaging capability of massive concurrent connection of iPush Server V2
- Performance of massive concurrent login of iPush Server V2
- Performance of massive concurrent subscription of iPush Server V2
Notice : There so many parameters that can affect the results of benchmarks. Any differences on benchmark scenarios, hardware, software and even network environments will make the benchmark results not the same.
In following benchmark reports, we use the same desktop as host :
| Model | Spec. |
| IBM M50 (8143I1V) |
|
Messaging capability of massive concurrent connection
| Server OS Tested | Windows 2000 Server (with iPush Server V2.1 Build105 Standalone for Windows) |
Here are summary of this benchmark result:
- None message dropped by iPush Server.
- (message size: 1KB) Throughput > 30,000 messages/sec. (Avg. CPU Usage : 50 %)(Avg. Network Usage : 26 %)
- (message size: 100 bytes.) Throughput > 105,000 messages/sec. ( Avg. CPU Usage : 45 %)(Avg. Network Usage : 13 %)
- The limitation of massive connection messaging capability of iPush Server has not been reached yet.
For further details please refer Benchmark_iPushV2_MassiveConnectionMessaging.pdf.
Performance of massive concurrent login handling
| Server OS Tested |
|
Here are summary of this benchmark result:
- Complete 3,000 simultaneous logins in 8.041 sec. (with CPU usage: 65%-75%, Avg. Memory Usage: 144 MB)
- The iPush Server versions for Windows or Linux get almost the same average time result.
- The bottleneck came from the DB (MySQL) responding to Auth. Center’s queries for user’s authentication and authorization data.
For further details please refer Benchmark_iPushV2_AuthCenter.pdf.
Performance of massive concurrent subscription handling
| Server OS Tested | Windows 2000 Server (with iPush Server V2.1 Build105 Standalone for Windows) |
Here are summary of this benchmark result:
- 100,000 subject subscriptions handled completely in 3.5 ~ 4 seconds by iPush Server.
- None subscription failed. Complete 28,377 subject subscriptions per second in average.
For further details please refer Benchmark_iPushV2_SubscriptionHandling.pdf.